Science
What happens when coffee roasts
Drying, browning, first crack and development — the chemistry behind roast character, at the depth that changes what a roaster does.
In short
A roast is usually described as a sequence of stages with neat borders. The borders are a convenience. What can be measured is a seed that dries, swells, loses a sixth or so of its weight, browns along a consistent path of colour and cracks audibly once or twice on the way.
This page keeps to what somebody measured or what the research literature reports, and says where the usual account runs ahead of the evidence. It gives no temperatures or times to aim for, because the sources do not support any.
The stages, as roasters name them
Sourced
A research group roasting on a commercial drum roaster divided each roast at three events an experienced roaster judged by eye and ear: the colour change, first crack and second crack.
Before the colour changes
The seed heats and dries. A thesis that measured aroma through the roast found little of it formed while the water content was still high, though it suggests this stage may matter for what forms later.
Colour change
The point at which the operator sees the beans have clearly left their original colour. It is a judgement, not an instrument reading.
Colour change to first crack
Browning proceeds. The same thesis found the fastest growth in aroma once the bean had dried below about five grams of water per hundred.
First crack
Audible popping begins. In the drum-roaster study, the acidity of the coffee reached its peak here in every profile tested.
First crack to second crack
Colour darkens further and measured acidity falls.
Second cracknot always done
A second, faster round of cracking. By this point the measured acidity had returned to the level of the green coffee.
Cooling
The roast is stopped by cooling the beans quickly, with air, with a measured spray of water, or both.
A descriptive model. The study that used these stages notes that the length of each one varies with how the roast is run, which is why no duration is shown.
Lighter and larger: what the bean loses and gains
Sourced
Coffee loses weight as it roasts. A doctoral thesis on roasting, summarising the earlier literature, gives the loss as generally between 14 and 20 per cent, depending on the green coffee, the process and how far the roast is taken. Most of that is water. A further part, put at some five to eight per cent for a medium roast, is dry matter that has turned to gas, chiefly carbon dioxide. The reactions themselves also produce water, which then leaves as vapour.
While it gets lighter the bean gets bigger. The same review reports one account of beans swelling by 40 to 60 per cent at a weight loss of 18 per cent, and notes that the account gave neither the starting moisture nor the roasting temperature. The thesis's own measurements found that, at the same degree of roast, beans roasted at higher temperature ended up larger and more porous than beans roasted more gently. Its author assumes, without having shown, that this more open structure lets aroma escape and staling proceed faster.
The practical consequence is that weight and volume move in opposite directions. A scoop of dark roast holds less coffee than a scoop of light, which is why a recipe measured by volume changes when the roast does.
A model of heat moving into one bean
CoffeeHQ explanation
A model of one bean, to explain why the outside and the inside of a roast are never at the same stage. The figures in it are two studies' measurements; the sequence is CoffeeHQ's.
Heat arrives at the surface
From hot air flowing past, from contact with hot metal and other beans, and by radiation from hot surfaces. The mix depends on the machine.
It moves inward slowly
Coffee is a poor conductor. One study measured the thermal conductivity of a mass of roasted beans at about a tenth of a watt per metre per degree, falling a little as the roast darkened, which is roughly a sixth of water's.
Water takes the heat first
While the bean is wet, much of the arriving heat goes into evaporating water, which holds the wet interior back. The thesis found beans with more water in them heated more slowly.
A gradient sets up
The surface runs ahead of the centre. A group that simulated a single bean in hot air describes the bean's temperature as, more precisely, a gradient from surface to centre, and the thesis measured a core cooler than the surrounding pile throughout.
Pressure builds, the structure gives
Steam and carbon dioxide form inside faster than they can escape. The bean swells, softens and cracks, and its pores open. No source read gives a figure for the pressure reached.
Lighter, larger, less dense
Mass falls while volume rises. The same study of thermal properties found the heat needed to warm a kilogram of beans by a degree rising as the roast darkened, from about 900 to about 1,250 joules.
A model. It explains why a very fast roast can be dark outside and under-roasted within, and why the same colour can hide different interiors; it does not say how large the gradient is in any roaster.
The reactions, as far as the literature commits itself
Sourced
Several families of reaction run at once: Maillard reactions between sugars and amino acids, the Strecker degradation of amino acids, caramelisation of sugars and, further on, pyrolysis. Between them they produce the aroma, flavour and colour of roasted coffee. The thesis says plainly that these reactions had not been completely worked out when it was written, because what happens inside a bean is hard to reproduce outside one. That caution is worth keeping: an account that assigns each flavour to one named reaction is tidier than the chemistry.
Some changes are better established. Chlorogenic acids break down in rough proportion to the degree of roast, and the review cites a loss of up to 80 per cent in dark roasts. Citric and malic acid fall, while quinic acid and the volatile acids rise. Caffeine is reported as thermally quite stable, trigonelline as partly degraded, and the oils as little affected.
Aroma does not simply accumulate. In the thesis's laboratory roasts some aroma compounds began to decline in the final stage as heat broke them down, while others went on increasing, so the mixture shifts in its last minutes as well as growing.
First and second crack: a landmark nobody has fully explained
Accounts differ
First crack is the event every roast is timed against, and its usual explanation is that steam and gas pressure burst the bean. The thesis that watched single beans through a microscope during roasting is more careful. It saw large cracks begin as a hairline on the flat face, which it says was probably accompanied by a sharp pop, and it suggests the popping sounds may be caused by escaping gas. Those are a researcher's hedged readings of what he saw, not a settled mechanism.
What is measured is the sound. An acoustics researcher recorded both and reported three differences: first crack is louder than second, it carries more low-frequency energy, and the cracks of second crack come at a faster rate. In the society's summary of the work he likens the first to popcorn and the second to a puffed-rice breakfast cereal. This rests on the paper's abstract and that summary; the full paper could not be opened for this page.
Two findings tie the cracks to something other than sound. In roasts run seven different ways on one drum roaster, the acidity of the coffee peaked at first crack every time and was back to its starting level by second crack. A second paper from the same group found the beans were approximately the same colour at first crack, and again at second crack, whichever way the roast had been run. In both, the moment of each crack was called by an experienced roaster, by ear and eye.
Does the roast start heating itself?
Accounts differ
It is often said that roasting turns exothermic near first crack: that the reactions begin to give out heat and the roast accelerates of its own accord. The literature is less sure than the saying. The thesis lists four sources that put the change from heat-absorbing to heat-releasing at 140, 150, 160 and 190 degrees Celsius, and then observes that the few temperature curves then published showed no rise in the final stage that could clearly be put down to it.
The idea is still current. The drum-roaster study describes a sudden late rise in a roast's rate of temperature increase as typically associated with the onset of exothermic reactions. The fair summary is that heat-releasing reactions are reported, that their onset has been placed across a range of fifty degrees, and that their visible effect on a roast curve is asserted more often than it has been demonstrated.
Cooling belongs to the roast
Sourced
The roast has to be stopped, and it is stopped by cooling the beans quickly. Industrially that is done with a large excess of cold air, with a measured amount of water sprayed on the hot beans, or with both. The water is meant to evaporate from the surface and carry heat away with it, not to raise the moisture of the coffee.
For anyone roasting at home this is the stage most often treated as tidying up. Two makers' manuals for home drum roasters instruct owners to let the machine's cooling cycle finish, and one tells owners never to unplug the machine during a roast.
Two things the colour of a roast does not tell you
Sourced
- How much caffeine is in the cup
- The older literature calls caffeine thermally quite stable. A study that roasted two coffees to many degrees and brewed them identically found something less tidy: caffeine in the cup was highest for light-to-medium roasts and lower for darker ones, partly because the darker roasts extracted less, and its authors read the pattern as indirect evidence that some caffeine is lost late in a roast. Where brews reached the same extraction, caffeine tended to rise with roast. Because dark beans also weigh less, measuring by scoop and by weight give different answers. No flat rule about roast and caffeine survives that.
- How the inside of the bean was roasted
- The thesis found that two roasts taken to the same degree by different temperatures differ in volume and pore structure, and its review notes that in very fast roasting colour is a poorer guide because the interior is less roasted than the surface. Two coffees can be the same colour and not the same coffee.
What this page does not give you
CoffeeHQ explanation
It gives no bean temperature for first or second crack, no duration for any stage and no target for weight loss. Figures of that kind circulate widely, but a temperature read by a probe in one machine is not the temperature of a bean, and the studies read for this page ran their roasts to suit their experiments, not to recommend a recipe.
It also does not say which reaction makes which flavour. The sources describe families of reaction and measured changes in particular compounds; none supports a simple table of cause and taste.
What to do next
- Coffee chemistry — The compounds behind the stages: what breaks down, what survives and what is made.
- Reading a roast profile — What a roast curve records, what rate of rise is, and how much of the advice about its shape has been tested.
- Coffee roasters and heat transfer — How heat reaches the bean in a drum and in an air roaster, and what that difference has been shown to do.
- Coffee roast defects — What baked, scorched and underdeveloped meant in the one study that made them on purpose.
- Coffee Expert level 7: roasting — The same material as a lesson, with exercises you can run on coffee you already have.